A Graph-Based Strategy for Intrusion Detection in Connected Vehicles

The CAN bus is the backbone for communication between various electronic control units (ECUs) in modern connected vehicles. However, the increasing connectivity and complexity of automotive systems have also introduced new security challenges, making the CAN bus vulnerable to intrusions and attacks. This paper proposes a method for detecting intrusions in the CAN bus using bidirected graphs. This study’s primary focus is building a mathematical model to identify anomalies using novel graph-based parameters like degree variance. We have chosen the dataset Car Hacking: Attack & Defense Challenge 2020 to test our proposed approach. We have achieved better accuracy in detecting attacks like DoS, fuzzy, spoofing, and replay. The proposed method performs well compared to existing techniques, notably when dealing with replay attacks, with the most remarkable accuracy of 98.38%. This method can detect and mitigate potential intrusions, ensuring connected vehicles’ safe and secure operation.